xref: /linux/drivers/accel/ivpu/ivpu_mmu.c (revision 263b2ba5fc93c875129e0d2b4034d7d8a34b3d39)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  * Copyright (C) 2020-2023 Intel Corporation
4  */
5 
6 #include <linux/circ_buf.h>
7 #include <linux/highmem.h>
8 
9 #include "ivpu_drv.h"
10 #include "ivpu_hw_mtl_reg.h"
11 #include "ivpu_hw_reg_io.h"
12 #include "ivpu_mmu.h"
13 #include "ivpu_mmu_context.h"
14 
15 #define IVPU_MMU_IDR0_REF		0x080f3e0f
16 #define IVPU_MMU_IDR0_REF_SIMICS	0x080f3e1f
17 #define IVPU_MMU_IDR1_REF		0x0e739d18
18 #define IVPU_MMU_IDR3_REF		0x0000003c
19 #define IVPU_MMU_IDR5_REF		0x00040070
20 #define IVPU_MMU_IDR5_REF_SIMICS	0x00000075
21 #define IVPU_MMU_IDR5_REF_FPGA		0x00800075
22 
23 #define IVPU_MMU_CDTAB_ENT_SIZE		64
24 #define IVPU_MMU_CDTAB_ENT_COUNT_LOG2	8 /* 256 entries */
25 #define IVPU_MMU_CDTAB_ENT_COUNT	((u32)1 << IVPU_MMU_CDTAB_ENT_COUNT_LOG2)
26 
27 #define IVPU_MMU_STREAM_ID0		0
28 #define IVPU_MMU_STREAM_ID3		3
29 
30 #define IVPU_MMU_STRTAB_ENT_SIZE	64
31 #define IVPU_MMU_STRTAB_ENT_COUNT	4
32 #define IVPU_MMU_STRTAB_CFG_LOG2SIZE	2
33 #define IVPU_MMU_STRTAB_CFG		IVPU_MMU_STRTAB_CFG_LOG2SIZE
34 
35 #define IVPU_MMU_Q_COUNT_LOG2		4 /* 16 entries */
36 #define IVPU_MMU_Q_COUNT		((u32)1 << IVPU_MMU_Q_COUNT_LOG2)
37 #define IVPU_MMU_Q_WRAP_BIT		(IVPU_MMU_Q_COUNT << 1)
38 #define IVPU_MMU_Q_WRAP_MASK		(IVPU_MMU_Q_WRAP_BIT - 1)
39 #define IVPU_MMU_Q_IDX_MASK		(IVPU_MMU_Q_COUNT - 1)
40 #define IVPU_MMU_Q_IDX(val)		((val) & IVPU_MMU_Q_IDX_MASK)
41 
42 #define IVPU_MMU_CMDQ_CMD_SIZE		16
43 #define IVPU_MMU_CMDQ_SIZE		(IVPU_MMU_Q_COUNT * IVPU_MMU_CMDQ_CMD_SIZE)
44 
45 #define IVPU_MMU_EVTQ_CMD_SIZE		32
46 #define IVPU_MMU_EVTQ_SIZE		(IVPU_MMU_Q_COUNT * IVPU_MMU_EVTQ_CMD_SIZE)
47 
48 #define IVPU_MMU_CMD_OPCODE		GENMASK(7, 0)
49 
50 #define IVPU_MMU_CMD_SYNC_0_CS		GENMASK(13, 12)
51 #define IVPU_MMU_CMD_SYNC_0_MSH		GENMASK(23, 22)
52 #define IVPU_MMU_CMD_SYNC_0_MSI_ATTR	GENMASK(27, 24)
53 #define IVPU_MMU_CMD_SYNC_0_MSI_ATTR	GENMASK(27, 24)
54 #define IVPU_MMU_CMD_SYNC_0_MSI_DATA	GENMASK(63, 32)
55 
56 #define IVPU_MMU_CMD_CFGI_0_SSEC	BIT(10)
57 #define IVPU_MMU_CMD_CFGI_0_SSV		BIT(11)
58 #define IVPU_MMU_CMD_CFGI_0_SSID	GENMASK(31, 12)
59 #define IVPU_MMU_CMD_CFGI_0_SID		GENMASK(63, 32)
60 #define IVPU_MMU_CMD_CFGI_1_RANGE	GENMASK(4, 0)
61 
62 #define IVPU_MMU_CMD_TLBI_0_ASID	GENMASK(63, 48)
63 #define IVPU_MMU_CMD_TLBI_0_VMID	GENMASK(47, 32)
64 
65 #define CMD_PREFETCH_CFG		0x1
66 #define CMD_CFGI_STE			0x3
67 #define CMD_CFGI_ALL			0x4
68 #define CMD_CFGI_CD			0x5
69 #define CMD_CFGI_CD_ALL			0x6
70 #define CMD_TLBI_NH_ASID		0x11
71 #define CMD_TLBI_EL2_ALL		0x20
72 #define CMD_TLBI_NSNH_ALL		0x30
73 #define CMD_SYNC			0x46
74 
75 #define IVPU_MMU_EVT_F_UUT		0x01
76 #define IVPU_MMU_EVT_C_BAD_STREAMID	0x02
77 #define IVPU_MMU_EVT_F_STE_FETCH	0x03
78 #define IVPU_MMU_EVT_C_BAD_STE		0x04
79 #define IVPU_MMU_EVT_F_BAD_ATS_TREQ	0x05
80 #define IVPU_MMU_EVT_F_STREAM_DISABLED	0x06
81 #define IVPU_MMU_EVT_F_TRANSL_FORBIDDEN	0x07
82 #define IVPU_MMU_EVT_C_BAD_SUBSTREAMID	0x08
83 #define IVPU_MMU_EVT_F_CD_FETCH		0x09
84 #define IVPU_MMU_EVT_C_BAD_CD		0x0a
85 #define IVPU_MMU_EVT_F_WALK_EABT	0x0b
86 #define IVPU_MMU_EVT_F_TRANSLATION	0x10
87 #define IVPU_MMU_EVT_F_ADDR_SIZE	0x11
88 #define IVPU_MMU_EVT_F_ACCESS		0x12
89 #define IVPU_MMU_EVT_F_PERMISSION	0x13
90 #define IVPU_MMU_EVT_F_TLB_CONFLICT	0x20
91 #define IVPU_MMU_EVT_F_CFG_CONFLICT	0x21
92 #define IVPU_MMU_EVT_E_PAGE_REQUEST	0x24
93 #define IVPU_MMU_EVT_F_VMS_FETCH	0x25
94 
95 #define IVPU_MMU_EVT_OP_MASK		GENMASK_ULL(7, 0)
96 #define IVPU_MMU_EVT_SSID_MASK		GENMASK_ULL(31, 12)
97 
98 #define IVPU_MMU_Q_BASE_RWA		BIT(62)
99 #define IVPU_MMU_Q_BASE_ADDR_MASK	GENMASK_ULL(51, 5)
100 #define IVPU_MMU_STRTAB_BASE_RA		BIT(62)
101 #define IVPU_MMU_STRTAB_BASE_ADDR_MASK	GENMASK_ULL(51, 6)
102 
103 #define IVPU_MMU_IRQ_EVTQ_EN		BIT(2)
104 #define IVPU_MMU_IRQ_GERROR_EN		BIT(0)
105 
106 #define IVPU_MMU_CR0_ATSCHK		BIT(4)
107 #define IVPU_MMU_CR0_CMDQEN		BIT(3)
108 #define IVPU_MMU_CR0_EVTQEN		BIT(2)
109 #define IVPU_MMU_CR0_PRIQEN		BIT(1)
110 #define IVPU_MMU_CR0_SMMUEN		BIT(0)
111 
112 #define IVPU_MMU_CR1_TABLE_SH		GENMASK(11, 10)
113 #define IVPU_MMU_CR1_TABLE_OC		GENMASK(9, 8)
114 #define IVPU_MMU_CR1_TABLE_IC		GENMASK(7, 6)
115 #define IVPU_MMU_CR1_QUEUE_SH		GENMASK(5, 4)
116 #define IVPU_MMU_CR1_QUEUE_OC		GENMASK(3, 2)
117 #define IVPU_MMU_CR1_QUEUE_IC		GENMASK(1, 0)
118 #define IVPU_MMU_CACHE_NC		0
119 #define IVPU_MMU_CACHE_WB		1
120 #define IVPU_MMU_CACHE_WT		2
121 #define IVPU_MMU_SH_NSH			0
122 #define IVPU_MMU_SH_OSH			2
123 #define IVPU_MMU_SH_ISH			3
124 
125 #define IVPU_MMU_CMDQ_OP		GENMASK_ULL(7, 0)
126 
127 #define IVPU_MMU_CD_0_TCR_T0SZ		GENMASK_ULL(5, 0)
128 #define IVPU_MMU_CD_0_TCR_TG0		GENMASK_ULL(7, 6)
129 #define IVPU_MMU_CD_0_TCR_IRGN0		GENMASK_ULL(9, 8)
130 #define IVPU_MMU_CD_0_TCR_ORGN0		GENMASK_ULL(11, 10)
131 #define IVPU_MMU_CD_0_TCR_SH0		GENMASK_ULL(13, 12)
132 #define IVPU_MMU_CD_0_TCR_EPD0		BIT_ULL(14)
133 #define IVPU_MMU_CD_0_TCR_EPD1		BIT_ULL(30)
134 #define IVPU_MMU_CD_0_ENDI		BIT(15)
135 #define IVPU_MMU_CD_0_V			BIT(31)
136 #define IVPU_MMU_CD_0_TCR_IPS		GENMASK_ULL(34, 32)
137 #define IVPU_MMU_CD_0_TCR_TBI0		BIT_ULL(38)
138 #define IVPU_MMU_CD_0_AA64		BIT(41)
139 #define IVPU_MMU_CD_0_S			BIT(44)
140 #define IVPU_MMU_CD_0_R			BIT(45)
141 #define IVPU_MMU_CD_0_A			BIT(46)
142 #define IVPU_MMU_CD_0_ASET		BIT(47)
143 #define IVPU_MMU_CD_0_ASID		GENMASK_ULL(63, 48)
144 
145 #define IVPU_MMU_CD_1_TTB0_MASK		GENMASK_ULL(51, 4)
146 
147 #define IVPU_MMU_STE_0_S1CDMAX		GENMASK_ULL(63, 59)
148 #define IVPU_MMU_STE_0_S1FMT		GENMASK_ULL(5, 4)
149 #define IVPU_MMU_STE_0_S1FMT_LINEAR	0
150 #define IVPU_MMU_STE_DWORDS		8
151 #define IVPU_MMU_STE_0_CFG_S1_TRANS	5
152 #define IVPU_MMU_STE_0_CFG		GENMASK_ULL(3, 1)
153 #define IVPU_MMU_STE_0_S1CTXPTR_MASK	GENMASK_ULL(51, 6)
154 #define IVPU_MMU_STE_0_V			BIT(0)
155 
156 #define IVPU_MMU_STE_1_STRW_NSEL1	0ul
157 #define IVPU_MMU_STE_1_CONT		GENMASK_ULL(16, 13)
158 #define IVPU_MMU_STE_1_STRW		GENMASK_ULL(31, 30)
159 #define IVPU_MMU_STE_1_PRIVCFG		GENMASK_ULL(49, 48)
160 #define IVPU_MMU_STE_1_PRIVCFG_UNPRIV	2ul
161 #define IVPU_MMU_STE_1_INSTCFG		GENMASK_ULL(51, 50)
162 #define IVPU_MMU_STE_1_INSTCFG_DATA	2ul
163 #define IVPU_MMU_STE_1_MEV		BIT(19)
164 #define IVPU_MMU_STE_1_S1STALLD		BIT(27)
165 #define IVPU_MMU_STE_1_S1C_CACHE_NC	0ul
166 #define IVPU_MMU_STE_1_S1C_CACHE_WBRA	1ul
167 #define IVPU_MMU_STE_1_S1C_CACHE_WT	2ul
168 #define IVPU_MMU_STE_1_S1C_CACHE_WB	3ul
169 #define IVPU_MMU_STE_1_S1CIR		GENMASK_ULL(3, 2)
170 #define IVPU_MMU_STE_1_S1COR		GENMASK_ULL(5, 4)
171 #define IVPU_MMU_STE_1_S1CSH		GENMASK_ULL(7, 6)
172 #define IVPU_MMU_STE_1_S1DSS		GENMASK_ULL(1, 0)
173 #define IVPU_MMU_STE_1_S1DSS_TERMINATE	0x0
174 
175 #define IVPU_MMU_REG_TIMEOUT_US		(10 * USEC_PER_MSEC)
176 #define IVPU_MMU_QUEUE_TIMEOUT_US	(100 * USEC_PER_MSEC)
177 
178 #define IVPU_MMU_GERROR_ERR_MASK ((REG_FLD(MTL_VPU_HOST_MMU_GERROR, CMDQ)) | \
179 				  (REG_FLD(MTL_VPU_HOST_MMU_GERROR, EVTQ_ABT)) | \
180 				  (REG_FLD(MTL_VPU_HOST_MMU_GERROR, PRIQ_ABT)) | \
181 				  (REG_FLD(MTL_VPU_HOST_MMU_GERROR, MSI_CMDQ_ABT)) | \
182 				  (REG_FLD(MTL_VPU_HOST_MMU_GERROR, MSI_EVTQ_ABT)) | \
183 				  (REG_FLD(MTL_VPU_HOST_MMU_GERROR, MSI_PRIQ_ABT)) | \
184 				  (REG_FLD(MTL_VPU_HOST_MMU_GERROR, MSI_ABT)))
185 
186 static char *ivpu_mmu_event_to_str(u32 cmd)
187 {
188 	switch (cmd) {
189 	case IVPU_MMU_EVT_F_UUT:
190 		return "Unsupported Upstream Transaction";
191 	case IVPU_MMU_EVT_C_BAD_STREAMID:
192 		return "Transaction StreamID out of range";
193 	case IVPU_MMU_EVT_F_STE_FETCH:
194 		return "Fetch of STE caused external abort";
195 	case IVPU_MMU_EVT_C_BAD_STE:
196 		return "Used STE invalid";
197 	case IVPU_MMU_EVT_F_BAD_ATS_TREQ:
198 		return "Address Request disallowed for a StreamID";
199 	case IVPU_MMU_EVT_F_STREAM_DISABLED:
200 		return "Transaction marks non-substream disabled";
201 	case IVPU_MMU_EVT_F_TRANSL_FORBIDDEN:
202 		return "MMU bypass is disallowed for this StreamID";
203 	case IVPU_MMU_EVT_C_BAD_SUBSTREAMID:
204 		return "Invalid StreamID";
205 	case IVPU_MMU_EVT_F_CD_FETCH:
206 		return "Fetch of CD caused external abort";
207 	case IVPU_MMU_EVT_C_BAD_CD:
208 		return "Fetched CD invalid";
209 	case IVPU_MMU_EVT_F_WALK_EABT:
210 		return " An external abort occurred fetching a TLB";
211 	case IVPU_MMU_EVT_F_TRANSLATION:
212 		return "Translation fault";
213 	case IVPU_MMU_EVT_F_ADDR_SIZE:
214 		return " Output address caused address size fault";
215 	case IVPU_MMU_EVT_F_ACCESS:
216 		return "Access flag fault";
217 	case IVPU_MMU_EVT_F_PERMISSION:
218 		return "Permission fault occurred on page access";
219 	case IVPU_MMU_EVT_F_TLB_CONFLICT:
220 		return "A TLB conflict";
221 	case IVPU_MMU_EVT_F_CFG_CONFLICT:
222 		return "A configuration cache conflict";
223 	case IVPU_MMU_EVT_E_PAGE_REQUEST:
224 		return "Page request hint from a client device";
225 	case IVPU_MMU_EVT_F_VMS_FETCH:
226 		return "Fetch of VMS caused external abort";
227 	default:
228 		return "Unknown CMDQ command";
229 	}
230 }
231 
232 static void ivpu_mmu_config_check(struct ivpu_device *vdev)
233 {
234 	u32 val_ref;
235 	u32 val;
236 
237 	if (ivpu_is_simics(vdev))
238 		val_ref = IVPU_MMU_IDR0_REF_SIMICS;
239 	else
240 		val_ref = IVPU_MMU_IDR0_REF;
241 
242 	val = REGV_RD32(MTL_VPU_HOST_MMU_IDR0);
243 	if (val != val_ref)
244 		ivpu_dbg(vdev, MMU, "IDR0 0x%x != IDR0_REF 0x%x\n", val, val_ref);
245 
246 	val = REGV_RD32(MTL_VPU_HOST_MMU_IDR1);
247 	if (val != IVPU_MMU_IDR1_REF)
248 		ivpu_dbg(vdev, MMU, "IDR1 0x%x != IDR1_REF 0x%x\n", val, IVPU_MMU_IDR1_REF);
249 
250 	val = REGV_RD32(MTL_VPU_HOST_MMU_IDR3);
251 	if (val != IVPU_MMU_IDR3_REF)
252 		ivpu_dbg(vdev, MMU, "IDR3 0x%x != IDR3_REF 0x%x\n", val, IVPU_MMU_IDR3_REF);
253 
254 	if (ivpu_is_simics(vdev))
255 		val_ref = IVPU_MMU_IDR5_REF_SIMICS;
256 	else if (ivpu_is_fpga(vdev))
257 		val_ref = IVPU_MMU_IDR5_REF_FPGA;
258 	else
259 		val_ref = IVPU_MMU_IDR5_REF;
260 
261 	val = REGV_RD32(MTL_VPU_HOST_MMU_IDR5);
262 	if (val != val_ref)
263 		ivpu_dbg(vdev, MMU, "IDR5 0x%x != IDR5_REF 0x%x\n", val, val_ref);
264 }
265 
266 static int ivpu_mmu_cdtab_alloc(struct ivpu_device *vdev)
267 {
268 	struct ivpu_mmu_info *mmu = vdev->mmu;
269 	struct ivpu_mmu_cdtab *cdtab = &mmu->cdtab;
270 	size_t size = IVPU_MMU_CDTAB_ENT_COUNT * IVPU_MMU_CDTAB_ENT_SIZE;
271 
272 	cdtab->base = dmam_alloc_coherent(vdev->drm.dev, size, &cdtab->dma, GFP_KERNEL);
273 	if (!cdtab->base)
274 		return -ENOMEM;
275 
276 	ivpu_dbg(vdev, MMU, "CDTAB alloc: dma=%pad size=%zu\n", &cdtab->dma, size);
277 
278 	return 0;
279 }
280 
281 static int ivpu_mmu_strtab_alloc(struct ivpu_device *vdev)
282 {
283 	struct ivpu_mmu_info *mmu = vdev->mmu;
284 	struct ivpu_mmu_strtab *strtab = &mmu->strtab;
285 	size_t size = IVPU_MMU_STRTAB_ENT_COUNT * IVPU_MMU_STRTAB_ENT_SIZE;
286 
287 	strtab->base = dmam_alloc_coherent(vdev->drm.dev, size, &strtab->dma, GFP_KERNEL);
288 	if (!strtab->base)
289 		return -ENOMEM;
290 
291 	strtab->base_cfg = IVPU_MMU_STRTAB_CFG;
292 	strtab->dma_q = IVPU_MMU_STRTAB_BASE_RA;
293 	strtab->dma_q |= strtab->dma & IVPU_MMU_STRTAB_BASE_ADDR_MASK;
294 
295 	ivpu_dbg(vdev, MMU, "STRTAB alloc: dma=%pad dma_q=%pad size=%zu\n",
296 		 &strtab->dma, &strtab->dma_q, size);
297 
298 	return 0;
299 }
300 
301 static int ivpu_mmu_cmdq_alloc(struct ivpu_device *vdev)
302 {
303 	struct ivpu_mmu_info *mmu = vdev->mmu;
304 	struct ivpu_mmu_queue *q = &mmu->cmdq;
305 
306 	q->base = dmam_alloc_coherent(vdev->drm.dev, IVPU_MMU_CMDQ_SIZE, &q->dma, GFP_KERNEL);
307 	if (!q->base)
308 		return -ENOMEM;
309 
310 	q->dma_q = IVPU_MMU_Q_BASE_RWA;
311 	q->dma_q |= q->dma & IVPU_MMU_Q_BASE_ADDR_MASK;
312 	q->dma_q |= IVPU_MMU_Q_COUNT_LOG2;
313 
314 	ivpu_dbg(vdev, MMU, "CMDQ alloc: dma=%pad dma_q=%pad size=%u\n",
315 		 &q->dma, &q->dma_q, IVPU_MMU_CMDQ_SIZE);
316 
317 	return 0;
318 }
319 
320 static int ivpu_mmu_evtq_alloc(struct ivpu_device *vdev)
321 {
322 	struct ivpu_mmu_info *mmu = vdev->mmu;
323 	struct ivpu_mmu_queue *q = &mmu->evtq;
324 
325 	q->base = dmam_alloc_coherent(vdev->drm.dev, IVPU_MMU_EVTQ_SIZE, &q->dma, GFP_KERNEL);
326 	if (!q->base)
327 		return -ENOMEM;
328 
329 	q->dma_q = IVPU_MMU_Q_BASE_RWA;
330 	q->dma_q |= q->dma & IVPU_MMU_Q_BASE_ADDR_MASK;
331 	q->dma_q |= IVPU_MMU_Q_COUNT_LOG2;
332 
333 	ivpu_dbg(vdev, MMU, "EVTQ alloc: dma=%pad dma_q=%pad size=%u\n",
334 		 &q->dma, &q->dma_q, IVPU_MMU_EVTQ_SIZE);
335 
336 	return 0;
337 }
338 
339 static int ivpu_mmu_structs_alloc(struct ivpu_device *vdev)
340 {
341 	int ret;
342 
343 	ret = ivpu_mmu_cdtab_alloc(vdev);
344 	if (ret) {
345 		ivpu_err(vdev, "Failed to allocate cdtab: %d\n", ret);
346 		return ret;
347 	}
348 
349 	ret = ivpu_mmu_strtab_alloc(vdev);
350 	if (ret) {
351 		ivpu_err(vdev, "Failed to allocate strtab: %d\n", ret);
352 		return ret;
353 	}
354 
355 	ret = ivpu_mmu_cmdq_alloc(vdev);
356 	if (ret) {
357 		ivpu_err(vdev, "Failed to allocate cmdq: %d\n", ret);
358 		return ret;
359 	}
360 
361 	ret = ivpu_mmu_evtq_alloc(vdev);
362 	if (ret)
363 		ivpu_err(vdev, "Failed to allocate evtq: %d\n", ret);
364 
365 	return ret;
366 }
367 
368 static int ivpu_mmu_reg_write(struct ivpu_device *vdev, u32 reg, u32 val)
369 {
370 	u32 reg_ack = reg + 4; /* ACK register is 4B after base register */
371 	u32 val_ack;
372 	int ret;
373 
374 	REGV_WR32(reg, val);
375 
376 	ret = REGV_POLL(reg_ack, val_ack, (val == val_ack), IVPU_MMU_REG_TIMEOUT_US);
377 	if (ret)
378 		ivpu_err(vdev, "Failed to write register 0x%x\n", reg);
379 
380 	return ret;
381 }
382 
383 static int ivpu_mmu_irqs_setup(struct ivpu_device *vdev)
384 {
385 	u32 irq_ctrl = IVPU_MMU_IRQ_EVTQ_EN | IVPU_MMU_IRQ_GERROR_EN;
386 	int ret;
387 
388 	ret = ivpu_mmu_reg_write(vdev, MTL_VPU_HOST_MMU_IRQ_CTRL, 0);
389 	if (ret)
390 		return ret;
391 
392 	return ivpu_mmu_reg_write(vdev, MTL_VPU_HOST_MMU_IRQ_CTRL, irq_ctrl);
393 }
394 
395 static int ivpu_mmu_cmdq_wait_for_cons(struct ivpu_device *vdev)
396 {
397 	struct ivpu_mmu_queue *cmdq = &vdev->mmu->cmdq;
398 
399 	return REGV_POLL(MTL_VPU_HOST_MMU_CMDQ_CONS, cmdq->cons, (cmdq->prod == cmdq->cons),
400 			 IVPU_MMU_QUEUE_TIMEOUT_US);
401 }
402 
403 static int ivpu_mmu_cmdq_cmd_write(struct ivpu_device *vdev, const char *name, u64 data0, u64 data1)
404 {
405 	struct ivpu_mmu_queue *q = &vdev->mmu->cmdq;
406 	u64 *queue_buffer = q->base;
407 	int idx = IVPU_MMU_Q_IDX(q->prod) * (IVPU_MMU_CMDQ_CMD_SIZE / sizeof(*queue_buffer));
408 
409 	if (!CIRC_SPACE(IVPU_MMU_Q_IDX(q->prod), IVPU_MMU_Q_IDX(q->cons), IVPU_MMU_Q_COUNT)) {
410 		ivpu_err(vdev, "Failed to write MMU CMD %s\n", name);
411 		return -EBUSY;
412 	}
413 
414 	queue_buffer[idx] = data0;
415 	queue_buffer[idx + 1] = data1;
416 	q->prod = (q->prod + 1) & IVPU_MMU_Q_WRAP_MASK;
417 
418 	ivpu_dbg(vdev, MMU, "CMD write: %s data: 0x%llx 0x%llx\n", name, data0, data1);
419 
420 	return 0;
421 }
422 
423 static int ivpu_mmu_cmdq_sync(struct ivpu_device *vdev)
424 {
425 	struct ivpu_mmu_queue *q = &vdev->mmu->cmdq;
426 	u64 val;
427 	int ret;
428 
429 	val = FIELD_PREP(IVPU_MMU_CMD_OPCODE, CMD_SYNC) |
430 	      FIELD_PREP(IVPU_MMU_CMD_SYNC_0_CS, 0x2) |
431 	      FIELD_PREP(IVPU_MMU_CMD_SYNC_0_MSH, 0x3) |
432 	      FIELD_PREP(IVPU_MMU_CMD_SYNC_0_MSI_ATTR, 0xf);
433 
434 	ret = ivpu_mmu_cmdq_cmd_write(vdev, "SYNC", val, 0);
435 	if (ret)
436 		return ret;
437 
438 	clflush_cache_range(q->base, IVPU_MMU_CMDQ_SIZE);
439 	REGV_WR32(MTL_VPU_HOST_MMU_CMDQ_PROD, q->prod);
440 
441 	ret = ivpu_mmu_cmdq_wait_for_cons(vdev);
442 	if (ret)
443 		ivpu_err(vdev, "Timed out waiting for consumer: %d\n", ret);
444 
445 	return ret;
446 }
447 
448 static int ivpu_mmu_cmdq_write_cfgi_all(struct ivpu_device *vdev)
449 {
450 	u64 data0 = FIELD_PREP(IVPU_MMU_CMD_OPCODE, CMD_CFGI_ALL);
451 	u64 data1 = FIELD_PREP(IVPU_MMU_CMD_CFGI_1_RANGE, 0x1f);
452 
453 	return ivpu_mmu_cmdq_cmd_write(vdev, "CFGI_ALL", data0, data1);
454 }
455 
456 static int ivpu_mmu_cmdq_write_tlbi_nh_asid(struct ivpu_device *vdev, u16 ssid)
457 {
458 	u64 val = FIELD_PREP(IVPU_MMU_CMD_OPCODE, CMD_TLBI_NH_ASID) |
459 		  FIELD_PREP(IVPU_MMU_CMD_TLBI_0_ASID, ssid);
460 
461 	return ivpu_mmu_cmdq_cmd_write(vdev, "TLBI_NH_ASID", val, 0);
462 }
463 
464 static int ivpu_mmu_cmdq_write_tlbi_nsnh_all(struct ivpu_device *vdev)
465 {
466 	u64 val = FIELD_PREP(IVPU_MMU_CMD_OPCODE, CMD_TLBI_NSNH_ALL);
467 
468 	return ivpu_mmu_cmdq_cmd_write(vdev, "TLBI_NSNH_ALL", val, 0);
469 }
470 
471 static int ivpu_mmu_reset(struct ivpu_device *vdev)
472 {
473 	struct ivpu_mmu_info *mmu = vdev->mmu;
474 	u32 val;
475 	int ret;
476 
477 	memset(mmu->cmdq.base, 0, IVPU_MMU_CMDQ_SIZE);
478 	clflush_cache_range(mmu->cmdq.base, IVPU_MMU_CMDQ_SIZE);
479 	mmu->cmdq.prod = 0;
480 	mmu->cmdq.cons = 0;
481 
482 	memset(mmu->evtq.base, 0, IVPU_MMU_EVTQ_SIZE);
483 	clflush_cache_range(mmu->evtq.base, IVPU_MMU_EVTQ_SIZE);
484 	mmu->evtq.prod = 0;
485 	mmu->evtq.cons = 0;
486 
487 	ret = ivpu_mmu_reg_write(vdev, MTL_VPU_HOST_MMU_CR0, 0);
488 	if (ret)
489 		return ret;
490 
491 	val = FIELD_PREP(IVPU_MMU_CR1_TABLE_SH, IVPU_MMU_SH_ISH) |
492 	      FIELD_PREP(IVPU_MMU_CR1_TABLE_OC, IVPU_MMU_CACHE_WB) |
493 	      FIELD_PREP(IVPU_MMU_CR1_TABLE_IC, IVPU_MMU_CACHE_WB) |
494 	      FIELD_PREP(IVPU_MMU_CR1_QUEUE_SH, IVPU_MMU_SH_ISH) |
495 	      FIELD_PREP(IVPU_MMU_CR1_QUEUE_OC, IVPU_MMU_CACHE_WB) |
496 	      FIELD_PREP(IVPU_MMU_CR1_QUEUE_IC, IVPU_MMU_CACHE_WB);
497 	REGV_WR32(MTL_VPU_HOST_MMU_CR1, val);
498 
499 	REGV_WR64(MTL_VPU_HOST_MMU_STRTAB_BASE, mmu->strtab.dma_q);
500 	REGV_WR32(MTL_VPU_HOST_MMU_STRTAB_BASE_CFG, mmu->strtab.base_cfg);
501 
502 	REGV_WR64(MTL_VPU_HOST_MMU_CMDQ_BASE, mmu->cmdq.dma_q);
503 	REGV_WR32(MTL_VPU_HOST_MMU_CMDQ_PROD, 0);
504 	REGV_WR32(MTL_VPU_HOST_MMU_CMDQ_CONS, 0);
505 
506 	val = IVPU_MMU_CR0_CMDQEN;
507 	ret = ivpu_mmu_reg_write(vdev, MTL_VPU_HOST_MMU_CR0, val);
508 	if (ret)
509 		return ret;
510 
511 	ret = ivpu_mmu_cmdq_write_cfgi_all(vdev);
512 	if (ret)
513 		return ret;
514 
515 	ret = ivpu_mmu_cmdq_write_tlbi_nsnh_all(vdev);
516 	if (ret)
517 		return ret;
518 
519 	ret = ivpu_mmu_cmdq_sync(vdev);
520 	if (ret)
521 		return ret;
522 
523 	REGV_WR64(MTL_VPU_HOST_MMU_EVTQ_BASE, mmu->evtq.dma_q);
524 	REGV_WR32(MTL_VPU_HOST_MMU_EVTQ_PROD_SEC, 0);
525 	REGV_WR32(MTL_VPU_HOST_MMU_EVTQ_CONS_SEC, 0);
526 
527 	val |= IVPU_MMU_CR0_EVTQEN;
528 	ret = ivpu_mmu_reg_write(vdev, MTL_VPU_HOST_MMU_CR0, val);
529 	if (ret)
530 		return ret;
531 
532 	val |= IVPU_MMU_CR0_ATSCHK;
533 	ret = ivpu_mmu_reg_write(vdev, MTL_VPU_HOST_MMU_CR0, val);
534 	if (ret)
535 		return ret;
536 
537 	ret = ivpu_mmu_irqs_setup(vdev);
538 	if (ret)
539 		return ret;
540 
541 	val |= IVPU_MMU_CR0_SMMUEN;
542 	return ivpu_mmu_reg_write(vdev, MTL_VPU_HOST_MMU_CR0, val);
543 }
544 
545 static void ivpu_mmu_strtab_link_cd(struct ivpu_device *vdev, u32 sid)
546 {
547 	struct ivpu_mmu_info *mmu = vdev->mmu;
548 	struct ivpu_mmu_strtab *strtab = &mmu->strtab;
549 	struct ivpu_mmu_cdtab *cdtab = &mmu->cdtab;
550 	u64 *entry = strtab->base + (sid * IVPU_MMU_STRTAB_ENT_SIZE);
551 	u64 str[2];
552 
553 	str[0] = FIELD_PREP(IVPU_MMU_STE_0_CFG, IVPU_MMU_STE_0_CFG_S1_TRANS) |
554 		 FIELD_PREP(IVPU_MMU_STE_0_S1CDMAX, IVPU_MMU_CDTAB_ENT_COUNT_LOG2) |
555 		 FIELD_PREP(IVPU_MMU_STE_0_S1FMT, IVPU_MMU_STE_0_S1FMT_LINEAR) |
556 		 IVPU_MMU_STE_0_V |
557 		 (cdtab->dma & IVPU_MMU_STE_0_S1CTXPTR_MASK);
558 
559 	str[1] = FIELD_PREP(IVPU_MMU_STE_1_S1DSS, IVPU_MMU_STE_1_S1DSS_TERMINATE) |
560 		 FIELD_PREP(IVPU_MMU_STE_1_S1CIR, IVPU_MMU_STE_1_S1C_CACHE_NC) |
561 		 FIELD_PREP(IVPU_MMU_STE_1_S1COR, IVPU_MMU_STE_1_S1C_CACHE_NC) |
562 		 FIELD_PREP(IVPU_MMU_STE_1_S1CSH, IVPU_MMU_SH_NSH) |
563 		 FIELD_PREP(IVPU_MMU_STE_1_PRIVCFG, IVPU_MMU_STE_1_PRIVCFG_UNPRIV) |
564 		 FIELD_PREP(IVPU_MMU_STE_1_INSTCFG, IVPU_MMU_STE_1_INSTCFG_DATA) |
565 		 FIELD_PREP(IVPU_MMU_STE_1_STRW, IVPU_MMU_STE_1_STRW_NSEL1) |
566 		 FIELD_PREP(IVPU_MMU_STE_1_CONT, IVPU_MMU_STRTAB_CFG_LOG2SIZE) |
567 		 IVPU_MMU_STE_1_MEV |
568 		 IVPU_MMU_STE_1_S1STALLD;
569 
570 	WRITE_ONCE(entry[1], str[1]);
571 	WRITE_ONCE(entry[0], str[0]);
572 
573 	clflush_cache_range(entry, IVPU_MMU_STRTAB_ENT_SIZE);
574 
575 	ivpu_dbg(vdev, MMU, "STRTAB write entry (SSID=%u): 0x%llx, 0x%llx\n", sid, str[0], str[1]);
576 }
577 
578 static int ivpu_mmu_strtab_init(struct ivpu_device *vdev)
579 {
580 	ivpu_mmu_strtab_link_cd(vdev, IVPU_MMU_STREAM_ID0);
581 	ivpu_mmu_strtab_link_cd(vdev, IVPU_MMU_STREAM_ID3);
582 
583 	return 0;
584 }
585 
586 int ivpu_mmu_invalidate_tlb(struct ivpu_device *vdev, u16 ssid)
587 {
588 	struct ivpu_mmu_info *mmu = vdev->mmu;
589 	int ret;
590 
591 	ret = mutex_lock_interruptible(&mmu->lock);
592 	if (ret)
593 		return ret;
594 
595 	if (!mmu->on) {
596 		ret = 0;
597 		goto unlock;
598 	}
599 
600 	ret = ivpu_mmu_cmdq_write_tlbi_nh_asid(vdev, ssid);
601 	if (ret)
602 		goto unlock;
603 
604 	ret = ivpu_mmu_cmdq_sync(vdev);
605 unlock:
606 	mutex_unlock(&mmu->lock);
607 	return ret;
608 }
609 
610 static int ivpu_mmu_cd_add(struct ivpu_device *vdev, u32 ssid, u64 cd_dma)
611 {
612 	struct ivpu_mmu_info *mmu = vdev->mmu;
613 	struct ivpu_mmu_cdtab *cdtab = &mmu->cdtab;
614 	u64 *entry;
615 	u64 cd[4];
616 	int ret;
617 
618 	if (ssid > IVPU_MMU_CDTAB_ENT_COUNT)
619 		return -EINVAL;
620 
621 	entry = cdtab->base + (ssid * IVPU_MMU_CDTAB_ENT_SIZE);
622 
623 	if (cd_dma != 0) {
624 		cd[0] = FIELD_PREP(IVPU_MMU_CD_0_TCR_T0SZ, 26) |
625 			FIELD_PREP(IVPU_MMU_CD_0_TCR_TG0, 0) |
626 			FIELD_PREP(IVPU_MMU_CD_0_TCR_IRGN0, 0) |
627 			FIELD_PREP(IVPU_MMU_CD_0_TCR_ORGN0, 0) |
628 			FIELD_PREP(IVPU_MMU_CD_0_TCR_SH0, 0) |
629 			FIELD_PREP(IVPU_MMU_CD_0_TCR_IPS, 3) |
630 			FIELD_PREP(IVPU_MMU_CD_0_ASID, ssid) |
631 			IVPU_MMU_CD_0_TCR_EPD1 |
632 			IVPU_MMU_CD_0_AA64 |
633 			IVPU_MMU_CD_0_R |
634 			IVPU_MMU_CD_0_ASET |
635 			IVPU_MMU_CD_0_V;
636 		cd[1] = cd_dma & IVPU_MMU_CD_1_TTB0_MASK;
637 		cd[2] = 0;
638 		cd[3] = 0x0000000000007444;
639 
640 		/* For global context generate memory fault on VPU */
641 		if (ssid == IVPU_GLOBAL_CONTEXT_MMU_SSID)
642 			cd[0] |= IVPU_MMU_CD_0_A;
643 	} else {
644 		memset(cd, 0, sizeof(cd));
645 	}
646 
647 	WRITE_ONCE(entry[1], cd[1]);
648 	WRITE_ONCE(entry[2], cd[2]);
649 	WRITE_ONCE(entry[3], cd[3]);
650 	WRITE_ONCE(entry[0], cd[0]);
651 
652 	clflush_cache_range(entry, IVPU_MMU_CDTAB_ENT_SIZE);
653 
654 	ivpu_dbg(vdev, MMU, "CDTAB %s entry (SSID=%u, dma=%pad): 0x%llx, 0x%llx, 0x%llx, 0x%llx\n",
655 		 cd_dma ? "write" : "clear", ssid, &cd_dma, cd[0], cd[1], cd[2], cd[3]);
656 
657 	ret = mutex_lock_interruptible(&mmu->lock);
658 	if (ret)
659 		return ret;
660 
661 	if (!mmu->on) {
662 		ret = 0;
663 		goto unlock;
664 	}
665 
666 	ret = ivpu_mmu_cmdq_write_cfgi_all(vdev);
667 	if (ret)
668 		goto unlock;
669 
670 	ret = ivpu_mmu_cmdq_sync(vdev);
671 unlock:
672 	mutex_unlock(&mmu->lock);
673 	return ret;
674 }
675 
676 static int ivpu_mmu_cd_add_gbl(struct ivpu_device *vdev)
677 {
678 	int ret;
679 
680 	ret = ivpu_mmu_cd_add(vdev, 0, vdev->gctx.pgtable.pgd_dma);
681 	if (ret)
682 		ivpu_err(vdev, "Failed to add global CD entry: %d\n", ret);
683 
684 	return ret;
685 }
686 
687 static int ivpu_mmu_cd_add_user(struct ivpu_device *vdev, u32 ssid, dma_addr_t cd_dma)
688 {
689 	int ret;
690 
691 	if (ssid == 0) {
692 		ivpu_err(vdev, "Invalid SSID: %u\n", ssid);
693 		return -EINVAL;
694 	}
695 
696 	ret = ivpu_mmu_cd_add(vdev, ssid, cd_dma);
697 	if (ret)
698 		ivpu_err(vdev, "Failed to add CD entry SSID=%u: %d\n", ssid, ret);
699 
700 	return ret;
701 }
702 
703 int ivpu_mmu_init(struct ivpu_device *vdev)
704 {
705 	struct ivpu_mmu_info *mmu = vdev->mmu;
706 	int ret;
707 
708 	ivpu_dbg(vdev, MMU, "Init..\n");
709 
710 	drmm_mutex_init(&vdev->drm, &mmu->lock);
711 	ivpu_mmu_config_check(vdev);
712 
713 	ret = ivpu_mmu_structs_alloc(vdev);
714 	if (ret)
715 		return ret;
716 
717 	ret = ivpu_mmu_strtab_init(vdev);
718 	if (ret) {
719 		ivpu_err(vdev, "Failed to initialize strtab: %d\n", ret);
720 		return ret;
721 	}
722 
723 	ret = ivpu_mmu_cd_add_gbl(vdev);
724 	if (ret) {
725 		ivpu_err(vdev, "Failed to initialize strtab: %d\n", ret);
726 		return ret;
727 	}
728 
729 	ret = ivpu_mmu_enable(vdev);
730 	if (ret) {
731 		ivpu_err(vdev, "Failed to resume MMU: %d\n", ret);
732 		return ret;
733 	}
734 
735 	ivpu_dbg(vdev, MMU, "Init done\n");
736 
737 	return 0;
738 }
739 
740 int ivpu_mmu_enable(struct ivpu_device *vdev)
741 {
742 	struct ivpu_mmu_info *mmu = vdev->mmu;
743 	int ret;
744 
745 	mutex_lock(&mmu->lock);
746 
747 	mmu->on = true;
748 
749 	ret = ivpu_mmu_reset(vdev);
750 	if (ret) {
751 		ivpu_err(vdev, "Failed to reset MMU: %d\n", ret);
752 		goto err;
753 	}
754 
755 	ret = ivpu_mmu_cmdq_write_cfgi_all(vdev);
756 	if (ret)
757 		goto err;
758 
759 	ret = ivpu_mmu_cmdq_write_tlbi_nsnh_all(vdev);
760 	if (ret)
761 		goto err;
762 
763 	ret = ivpu_mmu_cmdq_sync(vdev);
764 	if (ret)
765 		goto err;
766 
767 	mutex_unlock(&mmu->lock);
768 
769 	return 0;
770 err:
771 	mmu->on = false;
772 	mutex_unlock(&mmu->lock);
773 	return ret;
774 }
775 
776 void ivpu_mmu_disable(struct ivpu_device *vdev)
777 {
778 	struct ivpu_mmu_info *mmu = vdev->mmu;
779 
780 	mutex_lock(&mmu->lock);
781 	mmu->on = false;
782 	mutex_unlock(&mmu->lock);
783 }
784 
785 static void ivpu_mmu_dump_event(struct ivpu_device *vdev, u32 *event)
786 {
787 	u32 ssid = FIELD_GET(IVPU_MMU_EVT_SSID_MASK, event[0]);
788 	u32 op = FIELD_GET(IVPU_MMU_EVT_OP_MASK, event[0]);
789 	u64 fetch_addr = ((u64)event[7]) << 32 | event[6];
790 	u64 in_addr = ((u64)event[5]) << 32 | event[4];
791 	u32 sid = event[1];
792 
793 	ivpu_err(vdev, "MMU EVTQ: 0x%x (%s) SSID: %d SID: %d, e[2] %08x, e[3] %08x, in addr: 0x%llx, fetch addr: 0x%llx\n",
794 		 op, ivpu_mmu_event_to_str(op), ssid, sid, event[2], event[3], in_addr, fetch_addr);
795 }
796 
797 static u32 *ivpu_mmu_get_event(struct ivpu_device *vdev)
798 {
799 	struct ivpu_mmu_queue *evtq = &vdev->mmu->evtq;
800 	u32 idx = IVPU_MMU_Q_IDX(evtq->cons);
801 	u32 *evt = evtq->base + (idx * IVPU_MMU_EVTQ_CMD_SIZE);
802 
803 	evtq->prod = REGV_RD32(MTL_VPU_HOST_MMU_EVTQ_PROD_SEC);
804 	if (!CIRC_CNT(IVPU_MMU_Q_IDX(evtq->prod), IVPU_MMU_Q_IDX(evtq->cons), IVPU_MMU_Q_COUNT))
805 		return NULL;
806 
807 	clflush_cache_range(evt, IVPU_MMU_EVTQ_CMD_SIZE);
808 
809 	evtq->cons = (evtq->cons + 1) & IVPU_MMU_Q_WRAP_MASK;
810 	REGV_WR32(MTL_VPU_HOST_MMU_EVTQ_CONS_SEC, evtq->cons);
811 
812 	return evt;
813 }
814 
815 void ivpu_mmu_irq_evtq_handler(struct ivpu_device *vdev)
816 {
817 	u32 *event;
818 	u32 ssid;
819 
820 	ivpu_dbg(vdev, IRQ, "MMU event queue\n");
821 
822 	while ((event = ivpu_mmu_get_event(vdev)) != NULL) {
823 		ivpu_mmu_dump_event(vdev, event);
824 
825 		ssid = FIELD_GET(IVPU_MMU_EVT_SSID_MASK, event[0]);
826 		if (ssid != IVPU_GLOBAL_CONTEXT_MMU_SSID)
827 			ivpu_mmu_user_context_mark_invalid(vdev, ssid);
828 	}
829 }
830 
831 void ivpu_mmu_irq_gerr_handler(struct ivpu_device *vdev)
832 {
833 	u32 gerror_val, gerrorn_val, active;
834 
835 	ivpu_dbg(vdev, IRQ, "MMU error\n");
836 
837 	gerror_val = REGV_RD32(MTL_VPU_HOST_MMU_GERROR);
838 	gerrorn_val = REGV_RD32(MTL_VPU_HOST_MMU_GERRORN);
839 
840 	active = gerror_val ^ gerrorn_val;
841 	if (!(active & IVPU_MMU_GERROR_ERR_MASK))
842 		return;
843 
844 	if (REG_TEST_FLD(MTL_VPU_HOST_MMU_GERROR, MSI_ABT, active))
845 		ivpu_warn_ratelimited(vdev, "MMU MSI ABT write aborted\n");
846 
847 	if (REG_TEST_FLD(MTL_VPU_HOST_MMU_GERROR, MSI_PRIQ_ABT, active))
848 		ivpu_warn_ratelimited(vdev, "MMU PRIQ MSI ABT write aborted\n");
849 
850 	if (REG_TEST_FLD(MTL_VPU_HOST_MMU_GERROR, MSI_EVTQ_ABT, active))
851 		ivpu_warn_ratelimited(vdev, "MMU EVTQ MSI ABT write aborted\n");
852 
853 	if (REG_TEST_FLD(MTL_VPU_HOST_MMU_GERROR, MSI_CMDQ_ABT, active))
854 		ivpu_warn_ratelimited(vdev, "MMU CMDQ MSI ABT write aborted\n");
855 
856 	if (REG_TEST_FLD(MTL_VPU_HOST_MMU_GERROR, PRIQ_ABT, active))
857 		ivpu_err_ratelimited(vdev, "MMU PRIQ write aborted\n");
858 
859 	if (REG_TEST_FLD(MTL_VPU_HOST_MMU_GERROR, EVTQ_ABT, active))
860 		ivpu_err_ratelimited(vdev, "MMU EVTQ write aborted\n");
861 
862 	if (REG_TEST_FLD(MTL_VPU_HOST_MMU_GERROR, CMDQ, active))
863 		ivpu_err_ratelimited(vdev, "MMU CMDQ write aborted\n");
864 
865 	REGV_WR32(MTL_VPU_HOST_MMU_GERRORN, gerror_val);
866 }
867 
868 int ivpu_mmu_set_pgtable(struct ivpu_device *vdev, int ssid, struct ivpu_mmu_pgtable *pgtable)
869 {
870 	return ivpu_mmu_cd_add_user(vdev, ssid, pgtable->pgd_dma);
871 }
872 
873 void ivpu_mmu_clear_pgtable(struct ivpu_device *vdev, int ssid)
874 {
875 	ivpu_mmu_cd_add_user(vdev, ssid, 0); /* 0 will clear CD entry */
876 }
877